Agricultural remote sensing pollution early warning method and early warning system based on knowledge graph

Abstract

The invention discloses an agricultural remote sensing pollution early warning method and early warning system based on knowledge graph, which relates to the field of pollution early warning technology. The vegetation data in the monitoring area is identified and the vegetation coefficient is obtained by analysis. If the monitoring area is determined to be a pollution stress area according to the vegetation coefficient, the polluted area is screened out in the pollution stress area. The pollution risk assessment model after training determines the pollution risk level in the polluted area. When the pollution level exceeds the expectation, the pollution data in the polluted area and the risk early warning instruction are synchronously sent to the outside, and the risk early warning frequency is constrained. The trend degree is generated according to the change trend data of the vegetation coefficient Ets within the observation period. If the trend degree exceeds the expectation, the pollution control plan is given by the pollution control knowledge graph, and the pollution data after control is visualized. Targeted control of polluted areas can be achieved, and the pollution risk in the polluted areas can be alleviated and reduced.

Description

Technical Field

[0001] The present invention relates to the field of pollution early warning technology, and specifically to an agricultural remote sensing pollution early warning method and early warning system based on a knowledge graph. Background Art

[0002] Agricultural pollution early warning based on remote sensing is a technical means of using remote sensing technology to monitor and evaluate pollution generated in agricultural activities. Through remote sensing platforms such as satellites and drones, combined with radiation information in different bands such as visible light, infrared, and microwaves, agricultural pollution remote sensing can obtain multi-dimensional and multi-scale agricultural environmental information such as surface cover, vegetation conditions, and water quality. After processing and analysis, this information can reveal the distribution, degree, and changing trend of agricultural pollution problems such as pesticide residues, excessive use of fertilizers, and livestock and poultry manure pollution, providing a scientific basis for agricultural environmental protection and pollution control. Agricultural pollution remote sensing technology has the advantages of wide coverage, short monitoring cycle, and large amount of information. It is one of the important means of current agricultural environmental monitoring and assessment.

[0003] In the Chinese invention patent with application publication number CN116401557A, a remote sensing monitoring and early warning method and system for agricultural non-point source pollution is disclosed, which includes: obtaining remote sensing data of the agricultural area to be monitored; collecting geographic data of the agricultural area to be monitored; obtaining safety standards of the agricultural area to be monitored based on the geographic data; comparing and analyzing the remote sensing data with the safety standards to obtain the pollution results of the agricultural area to be monitored; and issuing an early warning when the pollution results fall into the early warning conditions. This method realizes remote sensing monitoring and early warning of agricultural non-point source pollution by collecting remote sensing data. The implementation method is simple, which reduces the difficulty of remote sensing monitoring and early warning of agricultural non-point source pollution. It does not require too many personnel to conduct on-site monitoring, has low labor costs, and high timeliness. It overcomes the defects of poor timeliness and human errors in human monitoring, and also solves the defects of agricultural non-point source pollution that are difficult to monitor and warn due to the wide temporal and spatial range, large uncertainty, and complex and difficult to control component processes of agricultural non-point source pollution.

[0004] Combined with the above application and the contents of the prior art:

[0005] When it is necessary to predict the soil pollution and water pollution status in the agricultural planting area, since the planting range of agricultural crops is usually relatively wide, when it is necessary to monitor or warn the pollution status in the corresponding area, if the sensor network is used for data collection, it may be difficult to monitor the pollution due to the small coverage of the sensor network and the high installation cost. Therefore, agricultural remote sensing imaging will be introduced when monitoring pollution in a large area;

[0006] When conducting pollution warnings based on agricultural remote sensing, especially when there are certain changes in the nutrient content and pollutant quality in the soil in the area, and when there are certain changes in the soil state, in this case, after obtaining and identifying the remote sensing image, although the pollution risk and polluted area can be identified, the degree of pollution cannot be effectively judged. This results in the management personnel being unable to make targeted treatments for the current pollution status after obtaining the risk warning.

[0007] To this end, the present invention provides an agricultural remote sensing pollution early warning method and early warning system based on knowledge graph. Summary of the invention

[0008] 1. Technical issues to be resolved

[0009] In view of the deficiencies of the prior art, the present invention provides an agricultural remote sensing pollution early warning method and early warning system based on knowledge graphs. By screening out polluted areas in the pollution stress area, the trained pollution risk assessment model determines the pollution risk level in the polluted area. When the pollution level exceeds expectations, the pollution data in the polluted area and the risk early warning instructions are synchronously sent to the outside, and the risk early warning frequency is constrained; the trend degree is generated according to the change trend data of the vegetation coefficient Ets within the observation period. If the trend degree exceeds expectations, the pollution control knowledge graph gives a pollution control plan, and the pollution data after control is visualized. Targeted control of polluted areas can be achieved, and the pollution risk in the polluted area can be alleviated and reduced, thereby solving the technical problems raised in the background technology.

[0010] (II) Technical solution

[0011] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0012] The agricultural remote sensing pollution early warning method based on knowledge graph includes: after collecting soil data in the monitoring area, generating a soil state coefficient Sot from the soil monitoring data; if the soil state coefficient Sot exceeds the state threshold, collecting data in the monitoring area, and aggregating the acquired environmental monitoring data to generate an environmental monitoring data set;

[0013] After preprocessing the multi-source data in the environmental monitoring data set, the quality of the preprocessed image data is analyzed, and the quality coefficient Qts is generated by the image quality index. After screening out low-quality images based on the quality coefficient Qts, the low-quality images are optimized; among them, after the resolution Ft, distortion St and peak signal-to-noise ratio Xq are treated as dimensionless, the corresponding weight coefficients are determined by referring to the hierarchical analysis method, and then the quality coefficient Qts is generated in the following way:

[0014]

[0015] The weight coefficient is, 0≤A f ≤1,0≤A s ≤1,0≤A x ≤1, C x is the constant correction factor;

[0016] Identify the vegetation data in the monitoring area and analyze and obtain the vegetation coefficient Ets. If the monitoring area is determined to be a pollution-stressed area based on the vegetation coefficient Ets, screen out the polluted area in the pollution-stressed area and obtain its area and boundary, and mark the polluted area on the electronic map;

[0017] The pollution risk level in the polluted area is determined by the trained pollution risk assessment model. When the pollution level exceeds expectations, the pollution data and risk warning instructions in the polluted area are synchronously sent to the outside, and the frequency of risk warnings is constrained;

[0018] The trend degree Qos is generated based on the changing trend data of the vegetation coefficient Ets during the observation period. If the trend degree Qos exceeds expectations, the pollution control knowledge map will give a pollution control plan, and after the polluted area is controlled, the pollution data in the polluted area will be visualized.

[0019] Furthermore, several monitoring points are set in the monitoring area, soil state data is monitored and acquired at the monitoring points, and the acquired soil state data is aggregated to generate a soil monitoring data set; a soil state coefficient Sot in the monitoring area is generated from the soil monitoring data set, and if the acquired soil state coefficient Sot exceeds a state threshold, a data collection instruction is issued to the outside;

[0020] After receiving the data collection instruction, multi-source remote sensing data is collected in the monitoring area, including multi-spectral and hyperspectral data in the monitoring area; a sensor network is placed in the monitoring area, and the sensor network collects environmental parameters in real time in the monitoring area; the acquired environmental monitoring data is aggregated to generate an environmental monitoring data set.

[0021] Furthermore, the soil moisture Som and pollutant content Pom are linearly normalized, and the corresponding data values ​​are mapped to the interval [0,1], and the soil state coefficient Sot is generated in the following way:

[0022]

[0023] Weight coefficient: 0≤k1≤1, 0≤k2≤1 and k2+k1=1; Som i is the soil moisture at the i-th monitoring point, Som b is the qualified standard value of soil moisture, Som a is the mean soil moisture; Pom iis the pollutant content at the i-th monitoring point, Pom b is the qualified standard value of pollutant content, Pom a is the mean value of pollutant content.

[0024] Furthermore, after acquiring the environmental monitoring data set, the timestamps of different data sources are synchronized so that the data collected at different times correspond to each other, and the data from different sources are spatially aligned so that the spatial positions of different data match. After completing the data alignment, the acquired data is preprocessed.

[0025] Further, the quality of the preprocessed image data is analyzed to obtain the corresponding quality index, and the quality coefficient Qts is generated from the image quality index; if the quality coefficient Qts is lower than the quality threshold, it is regarded as a low-quality image;

[0026] Several image enhancement schemes are collected and summarized to build an image enhancement scheme library to obtain the image quality characteristics of low-quality images. Based on the correspondence between the image quality characteristics and the image enhancement scheme, the image enhancement scheme library matches the image enhancement scheme for the low-quality image.

[0027] Furthermore, the monitoring area is divided into several sub-areas, and the optimized image data in each sub-area is obtained. After the remote sensing image data covering the sub-area is identified, the difference vegetation index is output. After the difference vegetation index in each sub-area is obtained, the vegetation coefficient Ets is generated in the following manner:

[0028]

[0029] Where n is the number of sub-regions, Rv i is the difference vegetation index in the ith sub-region, Rv a is the average value of the difference vegetation index, Rv b is the qualified target value of the difference vegetation index; weight coefficient: 0≤S1≤1, 0≤S2≤1, and S1+S2=1; if the obtained vegetation coefficient Ets is lower than the vegetation threshold, the corresponding sub-area will be determined as a pollution stress area.

[0030] Furthermore, after collecting relevant data in the pollution stress area, pollution analysis is performed on the pollution stress area to obtain soil pollution characteristics, water pollution characteristics, and pollution texture characteristics, and then a pollution feature set of the pollution stress area is generated;

[0031] The pollution stress area is gridded and each grid area is labeled with the corresponding pollution characteristics. Based on the feature labeling results, several grid areas are clustered using the trained clustering algorithm to screen out specific polluted areas in the pollution stress area.

[0032] Furthermore, the pollution data in the polluted area is used as input, and the pollution risk in the polluted area is evaluated by the trained pollution risk assessment model, and the pollution risk level is output after the pollution risk level is divided; the risk warning frequency is constrained, and the constraint method is as follows:

[0033]

[0034] Where n is the number of warning nodes for risk warning, T ij is the time interval from the ith warning node to the jth warning node, T a is the average value of the time interval, F is the pollution risk level; weight coefficient: 0≤S1≤1, 0≤S2≤1, and S1+S2=1.

[0035] Furthermore, remote sensing data in the polluted area is periodically obtained during the observation period, and the corresponding vegetation coefficient Ets is obtained at several consecutive time nodes. The time nodes before and after receiving the risk warning are sorted according to the time axis and correspond one to one. The trend degree Qos is calculated according to the degree of change of the vegetation coefficient Ets, as follows:

[0036]

[0037] Among them, Jt i is the trend median value of the i-th time node, Jt a is its mean value, Jt p is its qualified standard value;

[0038] i=1,2,...,k,k is the number of time nodes, and are the vegetation coefficients at the i-th time node before and after the risk warning, and is the corresponding mean, weight coefficient: 0≤S1≤1, 0≤S2≤1 and S2+S1=1;

[0039] If the trend degree Qos exceeds the trend threshold, a pollution control instruction will be issued to the outside; after receiving the pollution control instruction, soil and water pollution control will be used as the target word, and a pollution control knowledge graph will be built in advance.

[0040] The agricultural remote sensing pollution early warning system based on knowledge graph includes a soil monitoring unit. After collecting soil data in the monitoring area, the soil state coefficient Sot is generated from the soil monitoring data. If the soil state coefficient Sot exceeds the state threshold, data is collected in the monitoring area, and the acquired environmental monitoring data is aggregated to generate an environmental monitoring data set.

[0041] The multi-source data processing unit pre-processes the multi-source data in the environmental monitoring data set, analyzes the quality of the pre-processed image data, generates a quality coefficient Qts from the image quality index, and optimizes the low-quality images after screening out the low-quality images according to the quality coefficient Qts;

[0042] The pollution analysis unit identifies the vegetation data in the monitoring area and analyzes and obtains the vegetation coefficient Ets. If the monitoring area is determined to be a pollution-stressed area based on the vegetation coefficient Ets, the polluted area is screened out in the pollution-stressed area and its area and boundary are obtained, and the polluted area is marked on the electronic map;

[0043] The risk warning unit determines the pollution risk level in the polluted area by using the trained pollution risk assessment model. When the pollution level exceeds expectations, the pollution data in the polluted area and the risk warning instructions are sent to the outside simultaneously, and the frequency of risk warnings is constrained;

[0044] The pollution control unit generates a trend degree Qos based on the changing trend data of the vegetation coefficient Ets during the observation period. If the trend degree Qos exceeds expectations, the pollution control knowledge map will give a pollution control plan, and after the polluted area is controlled, the pollution data in the polluted area will be visualized.

[0045] (III) Beneficial effects

[0046] The present invention provides an agricultural remote sensing pollution early warning method and early warning system based on knowledge graph, which has the following beneficial effects:

[0047] 1. Evaluate and judge the current soil degradation risk in the region based on the soil status coefficient Sot. Through comprehensive analysis of the current soil status, we can timely understand the current potential pollution risk of the soil and its impact on the vegetation status; collect various data in the region through different channels and data sources, so that the analysis and early warning of pollution in the region can be more accurate and the coverage is wider.

[0048] 2. After acquiring the remote sensing image, the image quality is analyzed and the quality coefficient Qts is generated from the acquired quality analysis data. The image quality can be evaluated and screened based on the quality coefficient Qts, and low-quality images can be screened out to reduce the impact of low-quality images on pollution analysis results.

[0049] 3. Match the image enhancement scheme for low-quality images based on the image quality characteristics, achieve targeted optimization for low-quality images, and improve the quality of remote sensing images. After the vegetation coefficient Ets is generated from the vegetation data, the vegetation status in the area can be comprehensively evaluated to determine whether there is serious pollution in the monitoring area, and the polluted area can be preliminarily screened out.

[0050] 4. By clustering the pollution characteristics in the grid area, similar grid areas are merged according to the similarity between the pollution characteristics, so as to identify and determine the pollution areas in the monitoring area, so that the warning results can be clearer when conducting pollution warning.

[0051] 5. The trained pollution risk assessment model is used to assess the pollution risk in the polluted area. After identifying and extracting various pollution data in the polluted area, the pollution risk in the polluted area can be quickly rated, and managers can take corresponding measures according to the risk level.

[0052] 6. Constrain the current warning frequency based on the current pollution risk level, so that the warning frequency is consistent with the degree of pollution, ensure that the warning instructions issued to the outside can be fully received, and the pollution risk can be dealt with in a timely manner.

[0053] 7. Based on the changes in the vegetation coefficient Ets in the polluted area, the change in the current pollution status can be evaluated. If the pollution status in the polluted area gradually deteriorates, it can be quickly controlled. Based on the various pollution data in the polluted area, the pollution control knowledge map can provide a pollution control plan, which can achieve targeted control of the polluted area and alleviate and reduce the pollution risk in the polluted area. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a schematic diagram of the process of the agricultural remote sensing pollution early warning method based on the knowledge graph of the present invention;

[0055] Figure 2 The figure is a flow chart of the agricultural remote sensing pollution early warning method based on knowledge graph of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] See also Figure 1 The present invention provides an agricultural remote sensing pollution early warning method based on knowledge graph, including:

[0058] Step 1: After collecting soil data in the monitoring area, the soil state coefficient Sot is generated from the soil monitoring data. If the soil state coefficient Sot exceeds the state threshold, data is collected in the monitoring area, and the acquired environmental monitoring data is aggregated to generate an environmental monitoring data set;

[0059] The step 1 includes the following contents:

[0060] Step 101: After determining the pollution monitoring area, several monitoring points are set in the monitoring area to monitor and obtain soil moisture, structure and composition changes, as well as soil state data such as nutrient content and pollutant quality at the monitoring points, and the obtained soil state data are aggregated to generate a soil monitoring data set;

[0061] Step 102: Generate the soil state coefficient Sot in the monitoring area from the soil monitoring data set, wherein the soil moisture Som and the pollutant content Pom are linearly normalized and the corresponding data values ​​are mapped to the interval [0,1] in the following manner:

[0062]

[0063] Weight coefficient: 0≤k1≤1, 0≤k2≤1 and k2+k1=1; weight coefficient can be obtained by referring to the hierarchical analysis method; Som i is the soil moisture at the i-th monitoring point, Som b is the qualified standard value of soil moisture, Som a is the mean soil moisture; Pom i is the pollutant content at the i-th monitoring point, Pom b is the qualified standard value of pollutant content, Pom a is the mean value of pollutant content;

[0064] The status threshold is set in advance based on historical data and management expectations of soil degradation risks. If the soil status coefficient Sot obtained exceeds the status threshold, it means that there is a certain soil degradation risk in the monitoring area. At this time, a data collection instruction is issued to the outside.

[0065] When in use, when it is necessary to detect and identify the pollution status in a part of the area, first collect various soil data in the area, generate the soil status coefficient Sot from the soil data in the area, and evaluate and judge the current soil degradation risk in the area based on the soil status coefficient Sot. By comprehensively analyzing the current soil status, we can timely understand the current potential pollution risk of the soil and its impact on the vegetation status.

[0066] Step 103: after receiving the data collection instruction, multi-source remote sensing data is collected in the monitoring area, including multi-spectral and hyperspectral data in the monitoring area; a sensor network is placed in the monitoring area, specifically including soil, water quality and meteorological sensors, and the sensor network collects environmental parameters in real time in the monitoring area; the acquired environmental monitoring data is aggregated to generate an environmental monitoring data set;

[0067] When using, combine the contents in steps 101 to 103:

[0068] Various data in the area are collected through different channels and data sources, such as remote sensing imaging data and sensor data, so that the analysis and early warning of pollution in the area can be more accurate and cover a wider range.

[0069] Combined with the above application and the contents of the prior art:

[0070] When it is necessary to predict the soil pollution and water pollution status in the agricultural planting area, since the planting range of agricultural crops is usually relatively wide, when it is necessary to monitor or warn the pollution status in the corresponding area, if the sensor network is used for data collection, it may be difficult to monitor the pollution due to the small coverage of the sensor network and the high installation cost. Therefore, agricultural remote sensing imaging will be introduced when monitoring pollution in a large area;

[0071] When conducting pollution warnings based on agricultural remote sensing, especially when there are certain changes in the nutrient content and pollutant quality in the soil in the area, and when there are certain changes in the soil state, in this case, after obtaining and identifying the remote sensing image, although the pollution risk and polluted area can be identified, the degree of pollution cannot be effectively judged. This results in the management personnel being unable to make targeted treatments for the current pollution status after obtaining the risk warning.

[0072] Step 2: After preprocessing the multi-source data in the environmental monitoring data set, the quality of the preprocessed image data is analyzed, and the quality coefficient Qts is generated from the image quality index. After filtering out low-quality images based on the quality coefficient Qts, the low-quality images are optimized;

[0073] The step 2 includes the following contents:

[0074] Step 201: after acquiring the environmental monitoring data set, synchronize the timestamps of different data sources so that the data collected at different times correspond to each other, and spatially register the data from different sources so that the spatial positions of different data match each other. After completing the data registration, pre-process the acquired data.

[0075] Step 202: Analyze the quality of the pre-processed image data to obtain corresponding quality indicators, including resolution, color depth, distortion, peak signal-to-noise ratio, structural similarity, etc., and generate a quality coefficient Qts from the image quality indicators, wherein the resolution Ft, distortion St and peak signal-to-noise ratio Xq are treated as dimensionless, and the corresponding weight coefficients are determined by referring to the hierarchical analysis method, and then in the following manner:

[0076]

[0077] The weight coefficient is, 0≤A f ≤1,0≤A s ≤1,0≤A x ≤1, C x is a constant correction coefficient, and its value falls within [-0.15, 0.15];

[0078] A quality threshold is pre-set based on historical data and expectations for image quality management; if the quality coefficient Qts is lower than the quality threshold, it means that the corresponding image quality is low and it is regarded as a low-quality image;

[0079] When in use, the data quality is improved by preprocessing multi-source data. At the same time, after acquiring the remote sensing image, the image quality is analyzed and the quality coefficient Qts is generated from the acquired quality analysis data. The image quality can be evaluated and screened based on the quality coefficient Qts, thereby screening out low-quality images and reducing the impact of low-quality images on pollution analysis results.

[0080] Step 203: collect a number of image enhancement schemes and compile them into an image enhancement scheme library, obtain image quality features of low-quality images, match image enhancement schemes for low-quality images from the image enhancement scheme library according to the correspondence between the image quality features and the image enhancement schemes, and optimize the low-quality images with the image enhancement schemes; fuse the multi-source data that have been optimized and preprocessed;

[0081] When using, combine the contents in steps 201 to 203:

[0082] By identifying the image quality features of low-quality images and preparing an image enhancement solution library in advance, and matching image enhancement solutions for low-quality images based on the image quality features, targeted optimization can be achieved for low-quality images, thereby improving the quality of remote sensing images.

[0083] Step 3: Identify the vegetation data in the monitoring area and analyze and obtain the vegetation coefficient Ets. If the monitoring area is determined to be a pollution-stressed area based on the vegetation coefficient Ets, select the polluted area in the pollution-stressed area and obtain its area and boundary, and mark the polluted area on the electronic map;

[0084] The step three includes the following contents:

[0085] Step 301: divide the monitoring area into several sub-areas, obtain optimized image data in each sub-area, identify the remote sensing image data covering the sub-area and output the difference vegetation index, where the difference vegetation index is the difference between the reflectivity of the near infrared band and the red band, can better identify vegetation and water bodies, is sensitive to soil background changes, and is often used for agricultural monitoring;

[0086] After obtaining the difference vegetation index in each sub-area, the vegetation coefficient Ets is generated, and the vegetation health status is evaluated based on the vegetation coefficient as follows:

[0087]

[0088] Where n is the number of sub-regions, Rv i is the difference vegetation index in the ith sub-region, Rv a is the average value of the difference vegetation index, Rv b is the qualified target value of the difference vegetation index; weight coefficient: 0≤S1≤1, 0≤S2≤1, and S1+S2=1;

[0089] The vegetation threshold is set in advance based on historical data and the expected vegetation management in the polluted area. If the obtained vegetation coefficient Ets is lower than the vegetation threshold, it means that the vegetation state in the corresponding sub-area is poor, and there may be certain pollution and soil degradation. At this time, the corresponding sub-area is determined as a pollution stress area.

[0090] When in use, after acquiring remote sensing images, the vegetation data in the sub-area is extracted through image recognition. After the vegetation coefficient Ets is generated from the vegetation data, a comprehensive evaluation of the vegetation status in the area can be performed to determine whether there is serious pollution in the monitoring area, and the polluted area can be preliminarily screened out;

[0091] Step 302: after collecting relevant data in the pollution stress area, pollution analysis is performed on the pollution stress area, including using the collected hyperspectral data to perform soil spectrum analysis to obtain corresponding soil pollution characteristics; based on remote sensing data, water quality parameters in the pollution stress area are obtained, and after feature extraction, corresponding water quality pollution characteristics are obtained, such as TSS concentration, chlorophyll concentration, etc.;

[0092] For example, the TSS concentration can be calculated based on the relationship between the spectral characteristics of the water body and the TSS concentration;

[0093] TSS refers to the sum of all suspended particles and suspended liquids in the water, usually in milligrams per liter (mg / L). These suspended solids may include sediments, sludge, phytoplankton, algae, bacteria, etc., and their sources include natural processes (such as rock erosion and sediment deposition) and human activities (such as land development and sewage discharge in agriculture, industry and urbanization).

[0094] TSS is one of the important indicators for measuring the turbidity and pollution degree of water bodies. High levels of TSS concentration will affect the transparency, color, oxygen transfer rate and ecosystem of water bodies, and even lead to eutrophication of water bodies, obstruction of photosynthesis, sediment accumulation and damage to biological ecosystems. Extract the pollution texture features of the polluted area and obtain the pollution texture features in the sub-area. The texture features describe the spatial distribution and arrangement pattern between pixels in the local area of ​​the image, which is very useful for distinguishing different materials, surface cover types or pollution conditions.

[0095] Step 303: Summarize the pollution texture features, water pollution features and soil pollution features to generate a pollution feature set for the pollution stress area;

[0096] The pollution stress area is gridded, and each grid area is marked with corresponding pollution features; according to the feature marking results, a number of grid areas are clustered by a trained clustering algorithm, and specific polluted areas are screened out in the pollution stress area, and the area and boundary of the polluted area are obtained, and the polluted area is marked on the electronic map;

[0097] When using, combine the contents in steps 301 to 303:

[0098] After initially screening out the polluted areas, the pollution features in the grid areas are clustered and similar grid areas are merged based on the similarity between the pollution features. This allows the identification and determination of polluted areas within the monitoring area, thereby making the warning results clearer when conducting pollution warnings.

[0099] Step 4: The pollution risk level in the polluted area is determined by the trained pollution risk assessment model. When the pollution level exceeds expectations, the pollution data in the polluted area and the risk warning instructions are simultaneously sent to the outside, and the frequency of risk warnings is constrained;

[0100] The step 4 includes the following contents:

[0101] Step 401: Use a convolutional neural network to build an initial model, train and test the initial model using sample data, and obtain a trained pollution risk assessment model;

[0102] Taking the pollution data in the polluted area as input, such as the pollution characteristics, boundaries and area in the polluted area, the trained pollution risk assessment model assesses the pollution risk in the polluted area and outputs the pollution risk level after the pollution risk level is divided;

[0103] When in use, based on the acquisition of corresponding pollution data, the trained pollution risk assessment model will assess the pollution risk in the polluted area. After identifying and extracting various pollution data in the polluted area, it can quickly rate the pollution risk in the polluted area, and managers can take corresponding treatment measures based on the risk level.

[0104] Step 402: When the pollution level exceeds expectations, a risk warning instruction is issued to the outside, and the pollution data in the polluted area and the risk warning instruction are synchronously issued to the outside. The risk warning frequency is constrained in the following manner:

[0105]

[0106] Where n is the number of warning nodes for risk warning, T ij is the time interval from the ith warning node to the jth warning node, T a is the average value of the time interval, F is the pollution risk level; weight coefficient: 0≤S1≤1, 0≤S2≤1, and S1+S2=1; the value of the weight coefficient is consistent with the previous value;

[0107] Issue risk warning instructions and corresponding pollution data at warning nodes that meet the constraints;

[0108] When using, combine the contents in steps 401 to 402:

[0109] When it is necessary to issue a risk warning to the outside, the current warning frequency will be constrained according to the current pollution risk level, so that the warning frequency is adapted to the degree of pollution, ensuring that the warning instructions issued to the outside can be fully received and the pollution risk can be dealt with in a timely manner.

[0110] Step 5: Generate the trend degree Qos based on the change trend data of the vegetation coefficient Ets within the observation period. If the trend degree Qos exceeds expectations, the pollution control knowledge map will give a pollution control plan, and after the pollution area is controlled, the pollution data in the pollution area will be visualized;

[0111] The step five includes the following contents:

[0112] Step 501: Set a pollution observation period, periodically obtain remote sensing data in the polluted area during the observation period, and obtain the corresponding vegetation coefficient Ets at several consecutive time nodes. Sort the time nodes before and after receiving the risk warning according to the time axis and correspond one to one. Calculate the trend degree Qos according to the degree of change of the vegetation coefficient Ets in the following way:

[0113]

[0114] Among them, Jti is the trend median value of the i-th time node, Jt a is its mean value, Jt p is its qualified standard value;

[0115] i=1,2,…,k,k is the number of time nodes, and are the vegetation coefficients at the i-th time node before and after the risk warning, and is the corresponding mean, weight coefficient: 0≤S1≤1, 0≤S2≤1 and S2+S1=1; the value of the weight coefficient is consistent with the previous value;

[0116] Based on historical data and the expected trend of index changes, the trend threshold is set in advance; if the trend degree Qos exceeds the trend threshold, it means that the pollution risk in the current polluted area is still increasing. At this time, a pollution control instruction is issued to the outside;

[0117] When in use, by continuously monitoring and observing the current pollution status, the change of the current pollution status can be evaluated based on the change of the vegetation coefficient Ets in the polluted area. If the pollution status in the polluted area gradually deteriorates, it can be quickly controlled.

[0118] Step 502: After receiving the pollution control instruction, soil and water pollution control is used as the target word, and after deep search and entity relationship building, a pollution control knowledge graph is pre-built; using the trained matching model, according to the correspondence between pollution characteristics and pollution control solutions, a pollution control solution is given from the pollution control knowledge graph;

[0119] After executing the pollution control plan to control the polluted area, the pollution data and vegetation data in the polluted area are periodically obtained and visualized;

[0120] When using, combine the contents in steps 501 and 502:

[0121] When it is necessary to treat the polluted area, the pollution control knowledge graph can provide a pollution control plan based on the pollution data in the polluted area, so as to achieve targeted treatment of the polluted area and alleviate and reduce the pollution risk in the polluted area.

[0122] The method of constructing the knowledge graph of soil and water pollution control can refer to the following:

[0123] Clarify the purpose: for example, for environmental monitoring, pollution control plan formulation, policy effectiveness evaluation, etc. Set goals: determine the specific problems that the knowledge graph needs to solve, such as rapid identification of pollution sources, accurate recommendation of control technologies, etc.

[0124] Domain definition: clarify the types of soil and water pollution (such as heavy metal pollution, organic pollution, etc.) and treatment technologies (such as bioremediation, chemical treatment, etc.) that the knowledge graph will cover. Boundary determination: define the boundaries of the knowledge graph to avoid overly broad or redundant information.

[0125] Data collection and preprocessing:

[0126] Official data sources: such as monitoring data and governance reports released by environmental protection departments and water conservancy departments. Academic resources: obtain the latest research results and technological advances from academic journals and conference papers. Online data: collect public opinions and case sharing from news websites, social media, etc.

[0127] Data preprocessing, data cleaning: remove duplicate, erroneous and irrelevant data; formatting: convert data into a unified format for subsequent processing; text extraction: for unstructured data, use natural language processing technology (such as named entity recognition, relationship extraction) to extract entities, relationships and attributes.

[0128] Entity extraction uses rule-based, statistical model or deep learning methods to extract entities from text. Entity types include pollutant names, governance technologies, policies and regulations, geographic locations, etc.

[0129] Relationship extraction, relationship type: clarify the relationship type between entities, such as "pollutants-treatment technology". Extraction method: use dependency parsing, pattern matching or deep learning models to extract relationships.

[0130] Attribute extraction, attribute definition: define relevant attributes for entities, such as the chemical properties of pollutants, the efficiency of treatment technology, etc. Extraction process: combine text analysis and domain knowledge base to extract attribute values ​​from text.

[0131] Knowledge Fusion:

[0132] Entity alignment, method: use similarity calculation, clustering or entity linking techniques to align the same entities from different sources; Challenge: deal with variations in entity names, spelling errors, etc.

[0133] Relation fusion, conflict resolution: For conflicting relations from different sources, resolve conflicts by evaluating the reliability of data sources, timestamps, etc. Relation integration: Merge multiple relations into a unified relational representation.

[0134] Attribute integration, attribute value merging: For the same attribute, merge attribute values ​​from different sources to form a comprehensive entity description. Attribute value verification: Verify the accuracy and rationality of attribute values.

[0135] Data model construction:

[0136] Ontology design, concept definition: clarify the concepts, categories and attributes in the knowledge graph; hierarchy: design the hierarchy of concepts, such as "pollutants" can be subdivided into subcategories such as "heavy metal pollutants" and "organic pollutants".

[0137] Data model definition, entity type definition: define the attributes and relationships for each entity type. Constraints: define constraints on entities and relationships to ensure data accuracy and consistency.

[0138] Knowledge storage and query:

[0139] Knowledge storage: choose a suitable graph database (such as Neo4j) or RDF storage system to store knowledge graph data. Index optimization: create indexes for frequently queried entities and relationships to improve query efficiency.

[0140] Knowledge query, query interface: provide query interfaces such as RESTful API and GraphQL to support complex query requirements, query optimization: optimize query statements and query strategies to reduce query time and resource consumption.

[0141] The Analytic Hierarchy Process (AHP) is a decision-making method that decomposes decision-related elements into levels such as goals, criteria, and plans, and conducts qualitative and quantitative analysis on this basis. It is particularly suitable for dealing with target systems with hierarchical and staggered evaluation indicators, and when the target value is difficult to describe quantitatively, the AHP is an effective decision-making tool.

[0142] The core of the hierarchical analysis method is to decompose the decision problem into multiple levels to form a hierarchical structure, which usually includes the target level, the criterion level, the sub-criterion level and the solution level. By solving the eigenvector of the judgment matrix, the priority weight of each element of each level to an element of the previous level is obtained, and the final weight of each alternative solution to the total goal is recursively merged by the weighted sum method, so as to find the optimal solution.

[0143] See also Figure 2 The present invention provides an agricultural remote sensing pollution early warning system based on knowledge graph, including:

[0144] The soil monitoring unit, after collecting soil data in the monitoring area, generates a soil state coefficient Sot from the soil monitoring data. If the soil state coefficient Sot exceeds the state threshold, data is collected in the monitoring area, and the acquired environmental monitoring data is aggregated to generate an environmental monitoring data set;

[0145] The multi-source data processing unit pre-processes the multi-source data in the environmental monitoring data set, analyzes the quality of the pre-processed image data, generates a quality coefficient Qts from the image quality index, and optimizes the low-quality images after screening out the low-quality images according to the quality coefficient Qts;

[0146] The pollution analysis unit identifies the vegetation data in the monitoring area and analyzes and obtains the vegetation coefficient Ets. If the monitoring area is determined to be a pollution-stressed area based on the vegetation coefficient Ets, the polluted area is screened out in the pollution-stressed area and its area and boundary are obtained, and the polluted area is marked on the electronic map;

[0147] The risk warning unit determines the pollution risk level in the polluted area by using the trained pollution risk assessment model. When the pollution level exceeds expectations, the pollution data in the polluted area and the risk warning instructions are sent to the outside simultaneously, and the frequency of risk warnings is constrained;

[0148] The pollution control unit generates a trend degree Qos based on the changing trend data of the vegetation coefficient Ets during the observation period. If the trend degree Qos exceeds expectations, the pollution control knowledge map will give a pollution control plan, and after the polluted area is controlled, the pollution data in the polluted area will be visualized.

[0149] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0150] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0151] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0152] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0153] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0154] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0155] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.

[0156] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. The agricultural remote sensing pollution early warning method based on knowledge graph is characterized by: include, After collecting soil data in the monitoring area, the soil state coefficient is generated from the soil monitoring data. , if the soil state coefficient When the state threshold is exceeded, data is collected in the monitoring area, and the acquired environmental monitoring data is aggregated to generate an environmental monitoring data set; wherein, several monitoring points are set in the monitoring area, soil state data is acquired at the monitoring points, and the acquired soil state data is aggregated to generate a soil monitoring data set; the soil state coefficient in the monitoring area is generated from the soil monitoring data set , among which, soil moisture and pollutant content Perform linear normalization and map the corresponding data values ​​to the interval The soil state coefficient is generated as follows : Weight coefficient: , and ; is the soil moisture at the i-th monitoring point, is the qualified standard value of soil moisture, is the mean soil moisture; is the pollutant content at the i-th monitoring point, is the qualified standard value of pollutant content, is the mean value of pollutant content; Based on historical data and management expectations of soil degradation risks, state thresholds are set in advance; if the soil state coefficient obtained is When the state threshold is exceeded, a data collection instruction is issued to the outside; After receiving the data collection instruction, multi-source remote sensing data is collected in the monitoring area, including multi-spectral and hyperspectral data in the monitoring area; a sensor network is placed in the monitoring area, and the sensor network collects environmental parameters in real time in the monitoring area, and the acquired environmental monitoring data is aggregated to generate an environmental monitoring data set; After preprocessing the multi-source data in the environmental monitoring data set, the quality of the preprocessed image data is analyzed, and the quality coefficient is generated from the image quality index. , according to the quality coefficient After filtering out low-quality images, the low-quality images are optimized; among them, the resolution , Distortion Peak Signal-to-Noise Ratio After being treated as dimensionless, the corresponding weight coefficient is determined by referring to the hierarchical analysis method, and then the quality coefficient is generated as follows : The weight coefficient is, , , , is the constant correction factor; Identify vegetation data within the monitoring area and analyze to obtain vegetation coefficients , if based on the vegetation coefficient Determine the monitoring area as a pollution stress area, screen out the polluted area in the pollution stress area and obtain its area and boundary, and mark the polluted area on the electronic map; wherein, divide the monitoring area into several sub-areas, obtain the optimized image data in each sub-area, identify the remote sensing image data covering the sub-area and output the difference vegetation index, and generate the vegetation coefficient after obtaining the difference vegetation index in each sub-area , as follows: Where n is the number of sub-regions, is the difference vegetation index in the ith sub-region, is the average value of the difference vegetation index, is the qualified target value of the difference vegetation index; weight coefficient: , ,and ; If the vegetation coefficient obtained If the vegetation threshold is lower than the threshold, the corresponding sub-area will be identified as a pollution stress area; After collecting relevant data in the pollution stress area, conduct pollution analysis on the pollution stress area to obtain soil pollution characteristics, water pollution characteristics, pollution texture characteristics, and summarize to generate a pollution feature set of the pollution stress area; The pollution stress area is gridded and each grid area is marked with corresponding pollution features; based on the feature marking results, a number of grid areas are clustered using a trained clustering algorithm to screen out specific pollution areas in the pollution stress area; The pollution risk level in the polluted area is determined by the trained pollution risk assessment model. When the pollution level exceeds expectations, the pollution data and risk warning instructions in the polluted area are synchronously sent to the outside, and the frequency of risk warnings is constrained; According to the vegetation coefficient during the observation period The trend data of the change is generated by the trend degree , if the trend Exceeding expectations, the pollution control knowledge graph provides a pollution control plan, and after the polluted area is controlled, the pollution data in the polluted area is visualized.

2. The agricultural remote sensing pollution early warning method based on knowledge graph according to claim 1 is characterized in that: After acquiring the environmental monitoring data set, the timestamps of different data sources are synchronized so that the data collected at different times correspond to each other, and the data from different sources are spatially aligned so that the spatial positions of different data match. After completing the data alignment, the acquired data is preprocessed.

3. The agricultural remote sensing pollution early warning method based on knowledge graph according to claim 2 is characterized in that: Analyze the quality of the preprocessed image data, obtain the corresponding quality index, and generate the quality coefficient from the image quality index , if the quality coefficient If it is below the quality threshold, it is regarded as a low-quality image; Several image enhancement schemes are collected and summarized to build an image enhancement scheme library to obtain the image quality characteristics of low-quality images. Based on the correspondence between the image quality characteristics and the image enhancement scheme, the image enhancement scheme library matches the image enhancement scheme for the low-quality image.

4. The agricultural remote sensing pollution early warning method based on knowledge graph according to claim 3 is characterized by: Taking the pollution data in the polluted area as input, the trained pollution risk assessment model assesses the pollution risk in the polluted area and outputs the pollution risk level after the pollution risk level is divided; The frequency of risk warnings is restricted in the following ways: Among them, p is the number of warning nodes for risk warning, is the time interval from the i-th warning node to the j-th warning node, is the average value of the time interval, is the pollution risk level; weight coefficient: , ,and .

5. The agricultural remote sensing pollution early warning method based on knowledge graph according to claim 4 is characterized in that: Periodically obtain remote sensing data in the polluted area during the observation period, and obtain the corresponding vegetation coefficient at several consecutive time points , sort the time nodes before and after receiving the risk warning according to the time axis and correspond one by one, and Calculate the degree of change to obtain the trend , as follows: in, is the trend median value of the i-th time node, is its mean value, is its qualified standard value; , is the number of time nodes, and are the vegetation coefficients at the i-th time node before and after the risk warning, and is the corresponding mean, weight coefficient: , and ; If the trend When the trend threshold is exceeded, a pollution control instruction is issued to the outside. After receiving the pollution control instruction, soil and water pollution control is used as the target word, and a pollution control knowledge graph is pre-built.

6. An agricultural remote sensing pollution early warning system based on knowledge graph, applying the method described in any one of claims 1 to 5, characterized in that: include, The soil monitoring unit collects soil data in the monitoring area and generates soil state coefficients based on the soil monitoring data. , if the soil state coefficient When the state threshold is exceeded, data is collected in the monitoring area, and the acquired environmental monitoring data is aggregated to generate an environmental monitoring data set; The multi-source data processing unit pre-processes the multi-source data in the environmental monitoring data set, analyzes the quality of the pre-processed image data, and generates a quality coefficient based on the image quality index. , according to the quality coefficient After filtering out low-quality images, optimize the low-quality images; Pollution analysis unit, identifies vegetation data in the monitoring area and analyzes and obtains vegetation coefficients , if based on the vegetation coefficient Determine the monitoring area as a pollution stress area, screen out the polluted area in the pollution stress area and obtain its area and boundary, and mark the polluted area on the electronic map; The risk warning unit determines the pollution risk level in the polluted area by using the trained pollution risk assessment model. When the pollution level exceeds expectations, the pollution data in the polluted area and the risk warning instructions are sent to the outside simultaneously, and the frequency of risk warnings is constrained; Pollution control unit, based on the vegetation coefficient during the observation period The trend data of the change is generated by the trend degree , if the trend Exceeding expectations, the pollution control knowledge graph provides a pollution control plan, and after the polluted area is controlled, the pollution data in the polluted area is visualized.

Images